Task

human-versus-viral protein classification

This evaluation asks whether protein representations distinguish human and viral sequence labels. It is a classification task, not a direct test of immune function.

Sources

Protein Language Models Expose Viral Immune Mimicry · Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1

8 evaluations · 32 metric rows

At a glance

Inputs, training, access and other details

Explanatory profile: limited source coverage · Automated source review, 2026-09-16. Review applies to the cited claims; unresolved fields are listed below. Numerical results retain their own review status.

Data, procedure and scoring
Property	Description and evidence
Entity type	Paper-specific evaluation task; this profile is a descriptive evidence summary. Sources Protein Language Models Expose Viral Immune Mimicry · Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1
Datasets	Reviewed human and vertebrate-host viral protein records from Swiss-Prot/UniProtKB, with redundancy filtering described in Methods 2.1. Sources Protein Language Models Expose Viral Immune Mimicry · Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1
Organisms	Human proteins and proteins from viruses with a known vertebrate host. Sources Protein Language Models Expose Viral Immune Mimicry · Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1
Assays	Sequence-origin labels from curated database records. This classification endpoint is not an experimental immune-response assay. Sources Protein Language Models Expose Viral Immune Mimicry · Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1
Splits	Methods 2.1 assigns whole UniRef50 clusters to training or test sets. This profile records the split principle, not a verified membership manifest. Sources Protein Language Models Expose Viral Immune Mimicry · Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1
Allowed inputs	Protein sequence representations paired with the study’s human/viral labels. Sources Protein Language Models Expose Viral Immune Mimicry · Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1
Adaptation	Pretrained protein representations are evaluated through a study-specific classifier; a backbone name alone does not identify the full fitted pipeline. Sources Protein Language Models Expose Viral Immune Mimicry · Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1
Metrics	AUROC, log loss, accuracy, precision and recall are described; precision and recall use macro averaging. The linked result retains its original percentage unit. Sources Protein Language Models Expose Viral Immune Mimicry · Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1
Baselines	Table 1 compares the reported protein-representation configurations. They are classification comparators, not experimental immune-function controls. Sources Protein Language Models Expose Viral Immune Mimicry · Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1

How it works

How it worksConceptual assessment outline

Conceptual overview of the published statistical assessment.

Sources

Protein Language Models Expose Viral Immune Mimicry · Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1

What the evaluation establishes

The source uses reviewed protein records and separates training and test data by sequence clusters. That reduces direct overlap between related examples under the stated clustering rule. It reports classification metrics for different representations; classification errors and biological explanations of those errors are separate claims.

Sources

Protein Language Models Expose Viral Immune Mimicry · Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1

Evaluation design

Benchmarks bring together tasks and protocols. A task describes the biological question; a protocol defines a particular test.

Protocols

Held-out human-versus-virus protein classification (human-versus-viral protein classification)

These source-backed links do not make different protocols or scores interchangeable.

Published comparisons

Explore the results reported under one evaluation protocol. Each figure keeps its source, dataset and metric together; it is not a ranking across studies.

Comparison and metric

Held-out human-versus-virus protein classification · Table 1

AUROC (percent) · Higher values are better for this metric.

Compare protein origin classifiers; T 5 embeddings plus linear/tree models vs ESM 2 fine-tuning and simple controls. UniRef90 deduplication; proteins longer than1,600 residues excluded; UniRef50 clusters assigned80% training and20% test with no cluster shared. Separate four-fold error-analysis experiment not assigned toTable1.

Evaluation protocol · human and viral proteins

BL Length · Configuration · Author-reported evaluation61.97
AA n-grams · Configuration · Author-reported evaluation91.5
ESM2 8M · Configuration · Author-reported evaluation98.09
ESM2 35M · Configuration · Author-reported evaluation98.69
ESM2 150M · Configuration · Author-reported evaluation99.26
ESM2 650M · Configuration · Author-reported evaluation99.67
Linear-T5 · Configuration · Author-reported evaluation99.56
Tree-T5 · Configuration · Author-reported evaluation99.65

Source order is preserved. Plotted marks show point estimates; uncertainty, where reported, is retained in the printed values and table. Differences do not establish statistical significance.

Protein Language Models Expose Viral Immune Mimicry · Table 1: AUC (%), Held-out human-versus-virus protein classification

Values, uncertainty and evidence

AUROC: original source values
Tested entity	Printed value	Uncertainty	Evidence
BL Length · Configuration	61.97 percent	Not reported	Author-reported evaluation · source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row BL Length, column AUC (%); XML row2 column2
AA n-grams · Configuration	91.5 percent	Not reported	Author-reported evaluation · source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row AA n-grams, column AUC (%); XML row3 column2
ESM2 8M · Configuration	98.09 percent	Not reported	Author-reported evaluation · source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row ESM2 8M, column AUC (%); XML row4 column2
ESM2 35M · Configuration	98.69 percent	Not reported	Author-reported evaluation · source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row ESM2 35M, column AUC (%); XML row5 column2
ESM2 150M · Configuration	99.26 percent	Not reported	Author-reported evaluation · source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row ESM2 150M, column AUC (%); XML row6 column2
ESM2 650M · Configuration	99.67 percent	Not reported	Author-reported evaluation · source checkedProtein Language Models Expose Viral Immune Mimicry; Protein Language Models Expose Viral Immune Mimicry · Table 1, ESM2 650M row, AUC (%) column
Linear-T5 · Configuration	99.56 percent	Not reported	Author-reported evaluation · source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row Linear-T5, column AUC (%); XML row8 column2
Tree-T5 · Configuration	99.65 percent	Not reported	Author-reported evaluation · source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row Tree-T5, column AUC (%); XML row9 column2

Scope and limitations

Origin classification does not establish immune mimicry.
Table footnote says all values are percent but log-loss values are dimensionless; log-loss units quarantined.
Narrative n-gram AUC91.9 differs from Table1 value91.5; preserve table value.
No interval assigned unless printed in source cell.

Source transcription and grouping reviewed by automated source review on 2026-09-17. These experiments were not independently reproduced by rewire.

Tested entities and results

Release 2026-09-17-d277315f7d76 · 8 evaluations · 32 metric rows. Different protocols are not a single leaderboard.

Evaluation setupMetricEvidence origin

Results grouped by the exact reported evaluation
Metric and finding	Coverage and uncertainty	Evidence
ESM2 650M: human-versus-viral protein classification Configuration: ESM2 650MProtocol: Held-out human-versus-virus protein classification (human-versus-viral protein classification)Dataset: human and viral proteins Compare protein origin classifiers; T 5 embeddings plus linear/tree models vs ESM 2 fine-tuning and simple controls. UniRef90 deduplication; proteins longer than1,600 residues excluded; UniRef50 clusters assigned80% training and20% test with no cluster shared. Separate four-fold error-analysis experiment not assigned toTable1. Author-reported evaluation · Evaluation metadata: needs review
99.67% AUROC Unit: percent · Direction: higher	Uncertainty: not reported in legacy extract Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry; Protein Language Models Expose Viral Immune Mimicry · Table 1, ESM2 650M row, AUC (%) column Source checking is not independent reproduction.
96.85% Prec. Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row ESM2 650M, column Prec.; XML row7 column4 Source checking is not independent reproduction.
97.86% Accur. Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row ESM2 650M, column Accur.; XML row7 column3 Source checking is not independent reproduction.
96.68% Recall Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row ESM2 650M, column Recall; XML row7 column5 Source checking is not independent reproduction.
ESM2 8M: Held-out human-versus-virus protein classification Configuration: ESM2 8MProtocol: Held-out human-versus-virus protein classification (human-versus-viral protein classification)Dataset: human and viral proteins Compare protein origin classifiers; T 5 embeddings plus linear/tree models vs ESM 2 fine-tuning and simple controls. UniRef90 deduplication; proteins longer than1,600 residues excluded; UniRef50 clusters assigned80% training and20% test with no cluster shared. Separate four-fold error-analysis experiment not assigned toTable1. Author-reported evaluation · Evaluation metadata: needs review
92.15% Prec. Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row ESM2 8M, column Prec.; XML row4 column4 Source checking is not independent reproduction.
92.33% Recall Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row ESM2 8M, column Recall; XML row4 column5 Source checking is not independent reproduction.
98.09% AUROC Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row ESM2 8M, column AUC (%); XML row4 column2 Source checking is not independent reproduction.
AA n-grams: Held-out human-versus-virus protein classification Configuration: AA n-gramsProtocol: Held-out human-versus-virus protein classification (human-versus-viral protein classification)Dataset: human and viral proteins Compare protein origin classifiers; T 5 embeddings plus linear/tree models vs ESM 2 fine-tuning and simple controls. UniRef90 deduplication; proteins longer than1,600 residues excluded; UniRef50 clusters assigned80% training and20% test with no cluster shared. Separate four-fold error-analysis experiment not assigned toTable1. Author-reported evaluation · Evaluation metadata: needs review
91.5% AUROC Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row AA n-grams, column AUC (%); XML row3 column2 Source checking is not independent reproduction.
88.49% Prec. Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row AA n-grams, column Prec.; XML row3 column4 Source checking is not independent reproduction.
ESM2 35M: Held-out human-versus-virus protein classification Configuration: ESM2 35MProtocol: Held-out human-versus-virus protein classification (human-versus-viral protein classification)Dataset: human and viral proteins Compare protein origin classifiers; T 5 embeddings plus linear/tree models vs ESM 2 fine-tuning and simple controls. UniRef90 deduplication; proteins longer than1,600 residues excluded; UniRef50 clusters assigned80% training and20% test with no cluster shared. Separate four-fold error-analysis experiment not assigned toTable1. Author-reported evaluation · Evaluation metadata: needs review
98.69% AUROC Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row ESM2 35M, column AUC (%); XML row5 column2 Source checking is not independent reproduction.
93.81% Prec. Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row ESM2 35M, column Prec.; XML row5 column4 Source checking is not independent reproduction.
93.92% Recall Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row ESM2 35M, column Recall; XML row5 column5 Source checking is not independent reproduction.
95.83% Accur. Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row ESM2 35M, column Accur.; XML row5 column3 Source checking is not independent reproduction.
ESM2 150M: Held-out human-versus-virus protein classification Configuration: ESM2 150MProtocol: Held-out human-versus-virus protein classification (human-versus-viral protein classification)Dataset: human and viral proteins Compare protein origin classifiers; T 5 embeddings plus linear/tree models vs ESM 2 fine-tuning and simple controls. UniRef90 deduplication; proteins longer than1,600 residues excluded; UniRef50 clusters assigned80% training and20% test with no cluster shared. Separate four-fold error-analysis experiment not assigned toTable1. Author-reported evaluation · Evaluation metadata: needs review
95.54% Prec. Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row ESM2 150M, column Prec.; XML row6 column4 Source checking is not independent reproduction.
96.99% Accur. Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row ESM2 150M, column Accur.; XML row6 column3 Source checking is not independent reproduction.
95.48% Recall Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row ESM2 150M, column Recall; XML row6 column5 Source checking is not independent reproduction.
99.26% AUROC Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row ESM2 150M, column AUC (%); XML row6 column2 Source checking is not independent reproduction.
Tree-T5: Held-out human-versus-virus protein classification Configuration: Tree-T5Protocol: Held-out human-versus-virus protein classification (human-versus-viral protein classification)Dataset: human and viral proteins Compare protein origin classifiers; T 5 embeddings plus linear/tree models vs ESM 2 fine-tuning and simple controls. UniRef90 deduplication; proteins longer than1,600 residues excluded; UniRef50 clusters assigned80% training and20% test with no cluster shared. Separate four-fold error-analysis experiment not assigned toTable1. Author-reported evaluation · Evaluation metadata: needs review
97.7% Recall Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row Tree-T5, column Recall; XML row9 column5 Source checking is not independent reproduction.
97.7% Accur. Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row Tree-T5, column Accur.; XML row9 column3 Source checking is not independent reproduction.
99.65% AUROC Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row Tree-T5, column AUC (%); XML row9 column2 Source checking is not independent reproduction.
Linear-T5: Held-out human-versus-virus protein classification Configuration: Linear-T5Protocol: Held-out human-versus-virus protein classification (human-versus-viral protein classification)Dataset: human and viral proteins Compare protein origin classifiers; T 5 embeddings plus linear/tree models vs ESM 2 fine-tuning and simple controls. UniRef90 deduplication; proteins longer than1,600 residues excluded; UniRef50 clusters assigned80% training and20% test with no cluster shared. Separate four-fold error-analysis experiment not assigned toTable1. Author-reported evaluation · Evaluation metadata: needs review
97.57% Recall Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row Linear-T5, column Recall; XML row8 column5 Source checking is not independent reproduction.
97.57% Accur. Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row Linear-T5, column Accur.; XML row8 column3 Source checking is not independent reproduction.
BL Length: Held-out human-versus-virus protein classification Configuration: BL LengthProtocol: Held-out human-versus-virus protein classification (human-versus-viral protein classification)Dataset: human and viral proteins Compare protein origin classifiers; T 5 embeddings plus linear/tree models vs ESM 2 fine-tuning and simple controls. UniRef90 deduplication; proteins longer than1,600 residues excluded; UniRef50 clusters assigned80% training and20% test with no cluster shared. Separate four-fold error-analysis experiment not assigned toTable1. Author-reported evaluation · Evaluation metadata: needs review
78.5% Accur. Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row BL Length, column Accur.; XML row2 column3 Source checking is not independent reproduction.
78.5% Prec. Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row BL Length, column Prec.; XML row2 column4 Source checking is not independent reproduction.
78.5% Recall Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row BL Length, column Recall; XML row2 column5 Source checking is not independent reproduction.

Papers and result coverage

Last literature check: 2026-09-17. Primary-source discovery and table/protocol screening; source checked is not independently reproduced. Raw acquisitions not automatically numerical publication approval.

Paper or primary resource	Version	Reference
Protein Language Models Expose Viral Immune Mimicry	version of record	Read source DOI: 10.3390/v17091199

What is still missing

Independent batch review before import; preserve existing observation identities.

Search and extraction details

complete comparison tables extracted pending publication review

Searches

Protein Language Models Expose Viral Immune Mimicry primary paper benchmark results

Evidence locations

Table1; held-out protein-origin classification; separate error-analysis CV

Strengths and limitations

Strengths and considerations

Cluster separation makes the treatment of related sequences visible when interpreting classification accuracy.
Sources
Protein Language Models Expose Viral Immune Mimicry · Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1

Limitations and conditions

Distinguishing sequence origin does not establish an immune mechanism. Database selection, similarity filtering and label composition constrain generalisation.
Sources
Protein Language Models Expose Viral Immune Mimicry · Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1

Profile review details

Inspected the cited primary abstract and descriptive computational-evaluation sections. Review covers the descriptive claims shown; no executable protocol was reconstructed. Original numerical records retain their prior transcription review.

Stable record: reported-task-53506fe386e4a1

Evidence table

Inspect claims, sources and review details

Trace each statement to its source and review. A context-only reference supports the record generally; it does not verify an individual field. Source checking does not reproduce an experiment.

One row per statement and cited source. Multiple citations are not independent evaluations. Shared locators are labelled explicitly.

16 evidence rows matching the loaded filters

Claims, original sources and review scope · Release 2026-09-17-d277315f7d76
Property and statement	Original source and location	Review and provenance
Diagram caption Conceptual overview of the published statistical assessment. Individual claims	Protein Language Models Expose Viral Immune Mimicry Original source ↗ Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1 Version: version of record Retrieved: 2026-09-16T10:33:57.274Z	source checked automated source review · 2026-09-16 Audit details Inspected the cited primary abstract and descriptive computational-evaluation sections. Review covers the descriptive claims shown; no executable protocol was reconstructed. Original numerical records retain their prior transcription review. Field: `attributes.profile.diagram.caption` Source artifact SHA-256: `15250af2f75f70e2b6a3725d00bc7e276ed9d2d4a54f0ae9d6eaabf6be13e4a1` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Diagram steps ["Curated sequence-origin labels","Keep sequence clusters separate","Assess held-out classifications","Report classification metrics"] Individual claims	Protein Language Models Expose Viral Immune Mimicry Original source ↗ Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1 Version: version of record Retrieved: 2026-09-16T10:33:57.274Z	source checked automated source review · 2026-09-16 Audit details Inspected the cited primary abstract and descriptive computational-evaluation sections. Review covers the descriptive claims shown; no executable protocol was reconstructed. Original numerical records retain their prior transcription review. Field: `attributes.profile.diagram.steps` Source artifact SHA-256: `15250af2f75f70e2b6a3725d00bc7e276ed9d2d4a54f0ae9d6eaabf6be13e4a1` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Diagram title Conceptual assessment outline Individual claims	Protein Language Models Expose Viral Immune Mimicry Original source ↗ Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1 Version: version of record Retrieved: 2026-09-16T10:33:57.274Z	source checked automated source review · 2026-09-16 Audit details Inspected the cited primary abstract and descriptive computational-evaluation sections. Review covers the descriptive claims shown; no executable protocol was reconstructed. Original numerical records retain their prior transcription review. Field: `attributes.profile.diagram.title` Source artifact SHA-256: `15250af2f75f70e2b6a3725d00bc7e276ed9d2d4a54f0ae9d6eaabf6be13e4a1` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Entity type Paper-specific evaluation task; this profile is a descriptive evidence summary. Individual claims	Protein Language Models Expose Viral Immune Mimicry Original source ↗ Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1 Version: version of record Retrieved: 2026-09-16T10:33:57.274Z	source checked automated source review · 2026-09-16 Audit details Inspected the cited primary abstract and descriptive computational-evaluation sections. Review covers the descriptive claims shown; no executable protocol was reconstructed. Original numerical records retain their prior transcription review. Field: `attributes.profile.facts.0.value` Source artifact SHA-256: `15250af2f75f70e2b6a3725d00bc7e276ed9d2d4a54f0ae9d6eaabf6be13e4a1` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Datasets Reviewed human and vertebrate-host viral protein records from Swiss-Prot/UniProtKB, with redundancy filtering described in Methods 2.1. Individual claims	Protein Language Models Expose Viral Immune Mimicry Original source ↗ Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1 Version: version of record Retrieved: 2026-09-16T10:33:57.274Z	source checked automated source review · 2026-09-16 Audit details Inspected the cited primary abstract and descriptive computational-evaluation sections. Review covers the descriptive claims shown; no executable protocol was reconstructed. Original numerical records retain their prior transcription review. Field: `attributes.profile.facts.1.value` Source artifact SHA-256: `15250af2f75f70e2b6a3725d00bc7e276ed9d2d4a54f0ae9d6eaabf6be13e4a1` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Organisms Human proteins and proteins from viruses with a known vertebrate host. Individual claims	Protein Language Models Expose Viral Immune Mimicry Original source ↗ Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1 Version: version of record Retrieved: 2026-09-16T10:33:57.274Z	source checked automated source review · 2026-09-16 Audit details Inspected the cited primary abstract and descriptive computational-evaluation sections. Review covers the descriptive claims shown; no executable protocol was reconstructed. Original numerical records retain their prior transcription review. Field: `attributes.profile.facts.2.value` Source artifact SHA-256: `15250af2f75f70e2b6a3725d00bc7e276ed9d2d4a54f0ae9d6eaabf6be13e4a1` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Assays Sequence-origin labels from curated database records. This classification endpoint is not an experimental immune-response assay. Individual claims	Protein Language Models Expose Viral Immune Mimicry Original source ↗ Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1 Version: version of record Retrieved: 2026-09-16T10:33:57.274Z	source checked automated source review · 2026-09-16 Audit details Inspected the cited primary abstract and descriptive computational-evaluation sections. Review covers the descriptive claims shown; no executable protocol was reconstructed. Original numerical records retain their prior transcription review. Field: `attributes.profile.facts.3.value` Source artifact SHA-256: `15250af2f75f70e2b6a3725d00bc7e276ed9d2d4a54f0ae9d6eaabf6be13e4a1` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Splits Methods 2.1 assigns whole UniRef50 clusters to training or test sets. This profile records the split principle, not a verified membership manifest. Individual claims	Protein Language Models Expose Viral Immune Mimicry Original source ↗ Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1 Version: version of record Retrieved: 2026-09-16T10:33:57.274Z	source checked automated source review · 2026-09-16 Audit details Inspected the cited primary abstract and descriptive computational-evaluation sections. Review covers the descriptive claims shown; no executable protocol was reconstructed. Original numerical records retain their prior transcription review. Field: `attributes.profile.facts.4.value` Source artifact SHA-256: `15250af2f75f70e2b6a3725d00bc7e276ed9d2d4a54f0ae9d6eaabf6be13e4a1` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Allowed inputs Protein sequence representations paired with the study’s human/viral labels. Individual claims	Protein Language Models Expose Viral Immune Mimicry Original source ↗ Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1 Version: version of record Retrieved: 2026-09-16T10:33:57.274Z	source checked automated source review · 2026-09-16 Audit details Inspected the cited primary abstract and descriptive computational-evaluation sections. Review covers the descriptive claims shown; no executable protocol was reconstructed. Original numerical records retain their prior transcription review. Field: `attributes.profile.facts.5.value` Source artifact SHA-256: `15250af2f75f70e2b6a3725d00bc7e276ed9d2d4a54f0ae9d6eaabf6be13e4a1` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Adaptation Pretrained protein representations are evaluated through a study-specific classifier; a backbone name alone does not identify the full fitted pipeline. Individual claims	Protein Language Models Expose Viral Immune Mimicry Original source ↗ Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1 Version: version of record Retrieved: 2026-09-16T10:33:57.274Z	source checked automated source review · 2026-09-16 Audit details Inspected the cited primary abstract and descriptive computational-evaluation sections. Review covers the descriptive claims shown; no executable protocol was reconstructed. Original numerical records retain their prior transcription review. Field: `attributes.profile.facts.6.value` Source artifact SHA-256: `15250af2f75f70e2b6a3725d00bc7e276ed9d2d4a54f0ae9d6eaabf6be13e4a1` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact

All evidence (CSV)All evidence (JSONL)How to read the evidence

Sources and history

Release 2026-09-17-d277315f7d76 · Record review: needs review

2 source records and release history

Protein Language Models Expose Viral Immune Mimicry · Original source · version of record
Protein Language Models Expose Viral Immune Mimicry · Original source · version of record

Download this release

Technical metadata and extraction receipts

Stable ID: reported-task-53506fe386e4a1

areas: proteins-complexes
tasks: human-versus-viral protein classification
entity level: task
version: Not reported
task: human-versus-viral protein classification
scope note: Paper-specific evaluation task; protocol completeness requires further extraction.
comparison panels: id: part2-viral-immune-mimicry-2025-viruses-17-01199-t001-3a83e9bebb; title: Held-out human-versus-virus protein classification · Table 1; protocol id: paper-protocol-1352834b9391c1dacb; dataset id: reported-dataset-43f24c4dfb7351; metric: AUROC; unit: percent; direction: higher; result ids: paper-result-e2941183edbcf0f806; paper-result-088aaf71a44cdc5709; paper-result-a19e0533c00420b50c; paper-result-13ca62957ffa02eef1; paper-result-a541b9f3a77eec62e2; lit-b4-009; paper-result-b740d08b806206779f; paper-result-6536626a03a3ef62f0; source ids: part2-viral-immune-mimicry-2025; source locator: Table 1: AUC (%), Held-out human-versus-virus protein classification; context: Compare protein origin classifiers; T 5 embeddings plus linear/tree models vs ESM 2 fine-tuning and simple controls. UniRef90 deduplication; proteins longer than1,600 residues excluded; UniRef50 clusters assigned80% training and20% test with no cluster shared. Separate four-fold error-analysis experiment not assigned toTable1.; caveats: Origin classification does not establish immune mimicry.; Table footnote says all values are percent but log-loss values are dimensionless; log-loss units quarantined.; Narrative n-gram AUC91.9 differs from Table1 value91.5; preserve table value.; No interval assigned unless printed in source cell.; review: method: automated_source_review; date: 2026-09-17; id: part2-viral-immune-mimicry-2025-viruses-17-01199-t001-668a472b03; title: Held-out human-versus-virus protein classification · Table 1; protocol id: paper-protocol-1352834b9391c1dacb; dataset id: reported-dataset-43f24c4dfb7351; metric: Accur.; unit: percent; direction: higher; result ids: paper-result-91e7ac901444849ecd; paper-result-ca12e3e749c612d6d8; paper-result-ee7f4e8a864f523eaf; paper-result-544d2c1cd2bd46af9b; paper-result-32da435487ad9e4e7b; paper-result-7ffc4775000c61b1e4; paper-result-87b3b45c365edc42f3; paper-result-439ed50c8779070823; source ids: part2-viral-immune-mimicry-2025; source locator: Table 1: Accur., Held-out human-versus-virus protein classification; context: Compare protein origin classifiers; T 5 embeddings plus linear/tree models vs ESM 2 fine-tuning and simple controls. UniRef90 deduplication; proteins longer than1,600 residues excluded; UniRef50 clusters assigned80% training and20% test with no cluster shared. Separate four-fold error-analysis experiment not assigned toTable1.; caveats: Origin classification does not establish immune mimicry.; Table footnote says all values are percent but log-loss values are dimensionless; log-loss units quarantined.; Narrative n-gram AUC91.9 differs from Table1 value91.5; preserve table value.; No interval assigned unless printed in source cell.; review: method: automated_source_review; date: 2026-09-17; id: part2-viral-immune-mimicry-2025-viruses-17-01199-t001-e488608662; title: Held-out human-versus-virus protein classification · Table 1; protocol id: paper-protocol-1352834b9391c1dacb; dataset id: reported-dataset-43f24c4dfb7351; metric: Prec.; unit: percent; direction: higher; result ids: paper-result-a950e05fb8768ed4e8; paper-result-0c1a3836d475c5a506; paper-result-00f7697f28c706f26b; paper-result-1f9cd7bc0d3be4259f; paper-result-24e80b6223e7bd1109; paper-result-0c5e3578112cc7247a; paper-result-c583dfa48d139dadc8; paper-result-f67a4402bb1e2c96ba; source ids: part2-viral-immune-mimicry-2025; source locator: Table 1: Prec., Held-out human-versus-virus protein classification; context: Compare protein origin classifiers; T 5 embeddings plus linear/tree models vs ESM 2 fine-tuning and simple controls. UniRef90 deduplication; proteins longer than1,600 residues excluded; UniRef50 clusters assigned80% training and20% test with no cluster shared. Separate four-fold error-analysis experiment not assigned toTable1.; caveats: Origin classification does not establish immune mimicry.; Table footnote says all values are percent but log-loss values are dimensionless; log-loss units quarantined.; Narrative n-gram AUC91.9 differs from Table1 value91.5; preserve table value.; No interval assigned unless printed in source cell.; review: method: automated_source_review; date: 2026-09-17; id: part2-viral-immune-mimicry-2025-viruses-17-01199-t001-a4943fb2ad; title: Held-out human-versus-virus protein classification · Table 1; protocol id: paper-protocol-1352834b9391c1dacb; dataset id: reported-dataset-43f24c4dfb7351; metric: Recall; unit: percent; direction: higher; result ids: paper-result-b36e1a943ebc5dfb9a; paper-result-ccac699408a1229570; paper-result-4f2c34cb0e09f0167a; paper-result-4014edf9066fc226f1; paper-result-9d9d31080b47613e46; paper-result-a42799c550a1ae61e7; paper-result-67bb8fdb6692f4336f; paper-result-2aafdf0e911ae9cff5; source ids: part2-viral-immune-mimicry-2025; source locator: Table 1: Recall, Held-out human-versus-virus protein classification; context: Compare protein origin classifiers; T 5 embeddings plus linear/tree models vs ESM 2 fine-tuning and simple controls. UniRef90 deduplication; proteins longer than1,600 residues excluded; UniRef50 clusters assigned80% training and20% test with no cluster shared. Separate four-fold error-analysis experiment not assigned toTable1.; caveats: Origin classification does not establish immune mimicry.; Table footnote says all values are percent but log-loss values are dimensionless; log-loss units quarantined.; Narrative n-gram AUC91.9 differs from Table1 value91.5; preserve table value.; No interval assigned unless printed in source cell.; review: method: automated_source_review; date: 2026-09-17
benchmark research: review date: 2026-09-17; status: complete_comparison_tables_extracted_pending_publication_review; primary sources: part2-viral-immune-mimicry-2025; inspected locators: Table1; held-out protein-origin classification; separate error-analysis CV; searched queries: Protein Language Models Expose Viral Immune Mimicry primary paper benchmark results; gaps: Independent batch review before import; preserve existing observation identities.; claim scope: Primary-source discovery and table/protocol screening; source checked is not independently reproduced. Raw acquisitions not automatically numerical publication approval.
historical missing metadata: protocol version: not_reported_in_legacy_extract; split: not_reported_in_legacy_extract
metadata review scope: historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.
legacy kinds: benchmark
entity classification: review date: 2026-09-17; rationale: This source-scoped record identifies the biological prediction task and holds its paper context. Preserve the existing task identity; exact split, model adaptation and scoring remain in linked evaluations or separate protocol records.; source ids: viral-immune-mimicry-2025; source locator: Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1; ambiguities: A paper- or suite-specific task may constrain some inputs or metrics; that alone does not make it interchangeable with a complete versioned protocol. No protocol equivalence is inferred.; Some legacy profile Entity type facts use the generic phrase computational evaluation protocol. That boilerplate is not sufficient to establish a single fixed protocol identity or to merge this task with another protocol record.

Related records

dataset: human and viral proteins
evaluates task: Held-out human-versus-virus protein classification (human-versus-viral protein classification)

Suggest a correction